ANALYSIS OF ELECTRO STATIC DISCHARGE ON GAAS

Transcription

ANALYSIS OF ELECTRO STATIC DISCHARGE ON GAAS
Progress In Electromagnetics Research C, Vol. 22, 179–193, 2011
ANALYSIS OF ELECTRO STATIC DISCHARGE ON
GAAS-BASED LOW NOISE AMPLIFIER
C.-H. Kim1 , S.-M. Hwang1, * , and J.-H. Choi2
1 Reliability
Technology Research Center, Korea Electronics Technology Institute (KETI), Republic of Korea
2 Department
of Radio Engineering, Hanyang University, Republic of
Korea
Abstract—This paper studies static effect of communication Low
Noise Amplifier (LNA) that utilizes GaAs wafer.
It analyzes
the Electro-Static Discharge (ESD) effect, which occurs within
communication components, such as GaAs LNA, and describes testing
standard and methods. In order to find out GaAs LNA’s susceptibility
to static, two well-recognized communication GaAs LNA IC models
were selected to be tested. Commercial program allowed measuring
of static energy inserted within LNA’s internal circuit by running a
simulation about static discharge of GaAs LNA. Then we analyzed
malfunctions caused by static and discussed about architectural
problem and improvement according to the test and simulation result,
from the perspective of GaAs LNA’s electro static discharge.
1. INTRODUCTION
As a substitution for silicon semiconductor, Gallium Arsenide (GaAs)
element is getting the spotlight today. GaAs is a compound of
gallium and arsenic, which is efficient for high-frequency circuit by
having faster operation speed and lower heat generation than silicon
semiconductor [1]. ESD is a well-known reliability aspect in Si
technologies, and it has been seriously addressed during last years in
many research papers [2]. It is generally believed that GaAs circuits
have a low susceptibility to ESD.
ESD is the most common cause of malfunction for low-powered
components, such as large scale integration. Among semiconductor
product failures, more than 50% of them are caused by ESD and
Received 27 April 2011, Accepted 24 June 2011, Scheduled 4 July 2011
* Corresponding author: Soon-Mi Hwang (asfara@keti.re.kr).
180
Kim, Hwang, and Choi
overvoltage. In the case of static discharge, electric charge transfer
happens instantly and results in dielectric breakdown or metallization
melt within semiconductor device from discharged voltage and induced
current [3]. Figure 1 illustrates common failure of integrated circuit,
and Table 1 shows ESD susceptibility of various electronic devices.
LNA is a component that amplifies the signal while lowering the
noise figure of high-frequency signal so that it is widely used for
telecommunication. The low-noise amplifiers that were selected as
the subjects of the reliability assessment were chip-on-board products
Figure 1. Common failure of integrated circuit [4, 5].
Table 1. ESD Susceptibility of electronic devices (to human body
ESD) [6, 7].
Device Type
VMOS
MOSFET
GaAsFET
EPROM
JFET
SAW
OP AMP
CMOS
Schottky diodes
Film resistors
Bipolar transistors
ECL
SCR
Schottky TTL
Range of ESD Susceptibility (V)
30 ∼ 1800
100 ∼ 200
100 ∼ 800
100
140 ∼ 7000
150 ∼ 500
190 ∼ 2500
250 ∼ 3000
300 ∼ 2500
300 ∼ 3000
380 ∼ 7000
500 ∼ 1500
680 ∼ 2500
1100 ∼ 2500
Progress In Electromagnetics Research C, Vol. 22, 2011
181
that have gallium arsenide semiconductor (GaAs Phemt) on a ceramic
substrate. High-frequency signal loss can be reduced by using an
amplifier transistor in bare chip and directly mounting the components
around the circuit connected with the bare chip transistor on a
ceramic substrate. Moreover, thermal stability and reliability, the
most vulnerable part of active modules, can be improved because bare
transistor chip is 10 times better in thermal conductivity than existing
plastic packaging type transistor and easier in heat sink dissipation to
ceramic substrate.
Despite its features, weak resistance to ESD causes product
malfunction and damage to the business. So high-frequency circuit
concepts as LNA need specific requirements for possible HF ESD
protection.
2. TEST STANDARD AND METHOD FOR ESD
2.1. Test Standard
An ESD event can be subdivided into HBM (Human Body
Model), MM (Machine Model), and CDM (Charged Device Model).
HBM resistance has been most widely used standard for ESD
evaluation [8, 9]. Including human body, electrical equivalent circuit of
discharge path has the form of double-exponential, and ESD waveform
is determined by human body’s charged capacitance and discharge
resistance. Figure 2 illustrates HBM’s equivalent circuit and waveform.
Rd
Cs
V
High-voltage
DC supply
I
Relay Discharge
electrode
4
EUT
distribute C and L Cs-150 pF
Rd=330 ohm
Return connection
Current (A)
Rch
3
2
1
0
10 20 30 40 50
60 70 80
Time (ns)
(a)
(b)
Figure 2. (a) HBM’s equivalent circuit, (b) HBM’s waveform.
As for HBM, the international test method standard for ESD
can be categorized into Component and System. Component testing
standard has MIL-Std 883 and EIA/JEDEC, while System testing
standard includes IEC61000-4-2 and ANSI C.63-16. Table 2 shows
the summary of current and proposed HBM ESD standards.
182
Kim, Hwang, and Choi
Table 2. Summary of current and proposed HBM ESD standards.
Category
For
Component
For System
Max. Voltage
(Pol. +/-)
2 kV ~ 8 kV
MIL-Std 883
EIA/JEDEC
0 ~ 8 kV
Test Method 5.1
IEC61000-4-2
0 ~ 8 kV (Contact)
0 ~ 15 kV(Air)
(former IEC801-2)
0 ~ 8 kV (Contact)
ANSI C.63-16
0 ~ 15 kV(Air)
Discharge
Network
100 pF/1500 Ω
100 pF/1500 Ω
150 pF/330 Ω
150 pF/330 Ω
2.2. Test Method
There are two discharge methods for ESD simulator, air discharge
mode and contact discharge mode. In air discharge mode, a spark
is formed between the tip and ground [10]. Most of the responses from
linear simulator and non-linear arcs determine discharge current. In
this case, the simulator current can be modeled using the impedance as
seen from ground plane into the discharge tip. This impedance can be
trans-formed in the time domain and convoluted with the non-linear
arc [11, 12]. This yields the discharge current. Fully modeling the
arc via differential equations in a time stepping algorithm is principle
possible, but require additional measures to avoid divergence, as the
ionization equations are highly sensitive to errors in the electric field
across the gap [13, 14].
Most ESD testing is done in contact mode. As the discharge
is initiated by a relay and as the impedances as seen from the relay
contacts are neither known, nor easy to measure, a different simulation
approach needs to be used.
It is not a straightforward task to simulate contact mode discharge,
although the system can be regarded as linear (provided the relay
switched more or less like ideal switch). In contact mode a capacitor
and some elements of the simulator are pre-charged. To initiate the
discharge a relay is closed.
In this paper, we carried out a ESD experiment according to
IEC61000-4-2 standard, in contact mode.
Progress In Electromagnetics Research C, Vol. 22, 2011
183
Figure 3. Testing figure.
Table 3. Test condition.
Test Standard
IEC61000-4-2
Test Condition
Discharge
330 Ω, 150 pF
Test Mode
Contact Discharge
Test Voltage
From 1 kV to 4 kV, 1 kV step
Test Polarity
Positive and Negative
Test Interval
Over 1 sec.
Number of Sample
5 ea/test level
Table 4. Test result.
A type
Test Result
Number of
of Sample
Test Voltage
ESD
Test
Voltage
1 kV
2 kV
3 kV
4 kV
5
5
5
5
Number
of Failure
0
5
5
5
B type
Number
Number
of Sample of Failure
5
0
5
0
5
5
5
5
3. TEST RESULT OF ESD
3.1. Test Result
ESD test was done on two communication GaAs LNA IC models
according to IEC 61000-4-2 standard, in contact mode. Testing
equipment was Mini Zap (MZ-15) of Key-Tek. Table 3 shows test
condition and Figure 3 illustrates the testing figure.
The measurement items are S11 (Input Return Loss), S22 (Output
Return Loss), S21 (Gain), Noise figure, IIP3 (Third-order intercept
point) and Power consumption. As a result, the failure of sample A
was reported at 2 kV, and failure of sample B was reported at 3 kV.
184
Kim, Hwang, and Choi
0
S 22 (dB)
S11 (dB)
0
−10
−20
−10
−20
−30
−30
830 840
850 860 870 880 890
Frequency (MHz)
830
840
(a)
850 860 870 880
Frequency (MHz)
890
(b)
5
4
N.F. (dB)
S21 (dB)
30
20
10
3
2
1
0
830 840
0
830
850 860 870 880 890
Frequency (MHz)
IIP 3 (dB)
30
20
10
0
830
850 860 870
Frequency (MHz)
880
890
(d)
Power consumption (W)
(c)
840
0.8
0.6
0.4
0.2
0.0
840
850 860 870 880
Frequency (MHz)
(e)
890
1
2
3
Test Volltage (V)
4
(f)
Figure 4. Measurement performance of Sample A according to
different applied voltage. (a) S11 , (b) S22 , (c) S21 , (d) noise figure,
(e) IIP3, (f) power consumption.
Progress In Electromagnetics Research C, Vol. 22, 2011
0
−10
S 22 (dB)
S 11 (dB)
0
185
−20
−10
−20
−30
−30
1950 2000
1950 2000 2050 2100 2150
Frequency (MHz)
2050 2100 2150
Frequency (MHz)
(a)
(b)
5
4
N.F. (dB)
S 21 (dB)
30
20
10
3
2
1
0
0
1950
2000 2050 2100 2150
Frequency (MHz)
1950 2000 2050 2100 2150
Frequency (MHz)
(d)
Power consumption (W)
(c)
IIP 3 (dB)
30
20
10
0
1950
2000 2050 2100
Frequency (MHz)
(e)
2150
0.8
0.6
0.4
0.2
0.0
1
2
3
Test Volltage (V)
4
(f)
Figure 5. Measurement performance of Sample B according to
different applied voltage. (a) S11 , (b) S22 , (c) S21 , (d) noise figure,
(e) IIP3, (f) power consumption.
186
Kim, Hwang, and Choi
(a)
(b)
Figure 6. ESD induced failure (at 4 kV HBM ESD impulse).
(a) Metallization burn-out, (b) line open.
(a)
(b)
Figure 7.
Modeling of GaAs LNA. (a) FLO/EMC model,
(b) consisting of LNA.
Table 4 shows the test result, and Figures 4 and 5 show measurement
performance of Samples A, B according to different applied voltages.
3.2. Failure Analysis
After the ESD test, the failed samples were carefully examined
with optical microscope and Environmental Scanning Electron
Microscope (ESEM), for further analysis. From the analysis, either
broken wire or burning trace could be observed in each sample.
Figure 6 illustrates ESD induced failure at 4 kV HBM ESD impulse.
4. SIMULATION OF ESD
4.1. Numerical Modeling
For further analysis on ESD test result of GaAs LNA, a
numerical modeling was implemented, using a commercial program,
Progress In Electromagnetics Research C, Vol. 22, 2011
187
FLO/EMC6.1 (Flomerics, Co., Ltd.) FLO/EMC’s basic modeling
approach is like the following [15, 16].
• Construct EMC model.
• Construct a Maxwell equation about the model and interpret 3D
vector field using TLM (Transmission-Line Matrix) method.
• Find a value from time-domain.
• Calculates vector field and magnetic field.
Figure 7 illustrates FLO/EMC model of GaAs LNA. The model has
two parts; ESD Generator and LNA device. LNA device consists of
ceramic substrate, metal plate, and package part.
Currnt (A)
4
3
2
1
0
0
10 20 30
40
50
60
70 80
Times (ns)
Figure 8. Input waveform of ESD (at 4 kV ESD impulse, defined in
IEC61000-4-2).
Electric Field (V/m)
4000
3000
2000
1000
0
(a)
0
5
10
15
Distance from Axis (mm)
20
(b)
Figure 9. Analysis of electric field distribution (at 4 kV ESD impulse).
(a) Electric field distribution, (b) electric field strength.
188
Kim, Hwang, and Choi
4.2. Simulation Result
Figure 8 illustrates and compares among input waveform of contact
discharge 4 kV, defined in IEC61000-4-2, the actual waveform
measured from GaAs LNA static discharge experiment, and input
waveform from static discharge test modeling. The waveform from the
actual experiment and the model are almost identical to the standard
waveform. Figure 8 compares the accuracy of electro static wave which
would be entered on testing sample. (This wave is not the one that
passes through LNA). Simulated wave has a similar overall shape to
that of the ideal wave but yet, not identical. Every time when a
measurement is taken, there exists a slight difference. It seems to
be a human or measurement error.
Simulation result suggests that LNA input circuit’s energy rises
as applied static level increases. Figure 9 illustrates analysis of electric
field distribution, which occurred within LNA’s circuit during the static
discharge experiment. When 4 kV ESD was applied, maximum of
36,000 V/m electric field was distributed throughout the entire circuit.
5. ESD SOLUTION FOR GAAS LNA
LNA failure caused by ESD show disconnection or burning in LNA
module’s input/output port. This might occur during production or
transportation process when a worker’s accumulated static instantly
flows into the product.
It happens more frequently for GaAs wafer than silicon wafer
because GaAs is weak at statics. The best solution would be to increase
its resistance to static, however, its physical property make it difficult.
Hence this thesis took the next best solution to improve the circuit.
To solve a case where ESD current harms the circuit through
Vee
C3
C4
R1
L3
D1 C 1 L 2
RF IN
C2
C5
R3
RF OUT
R2
L1
Input Matching
circuit
Input Matching
circuit
(a)
Figure 10.
(b) LNA.
LNA with protective diode.
(b)
(a) Matching circuit,
Progress In Electromagnetics Research C, Vol. 22, 2011
0
S 22 (dB)
0
S 11 (dB)
189
−10
−20
−30
−10
−20
−30
830 840
850
860
870
880
890
830 840
850
860
870
Frequency (MHz)
Frequency (MHz)
(a)
(b)
880
890
880
890
5
4
N.F. (dB)
S 21 (dB)
20
10
0
840
850
860
870
880
IIP 3 (dB)
0
830
890
840
850
860
870
Frequency (MHz)
Frequency (MHz)
(c)
(d)
30
20
10
0
830
2
1
Power consumption (W)
830
3
0.8
0.6
0.4
0.2
0.0
840
850 860
870
880
890
1
2
3
Frequency (MHz)
Test Volltage (V)
(e)
(f)
4
Figure 11.
Measurement performance of improved Sample A
according to different applied voltage. (a) S11 , (b) S22 , (c) S21 ,
(d) noise figure, (e) IIP3, (f) power consumption.
Kim, Hwang, and Choi
0
0
−10
−10
S 22 (dB)
S 11 (dB)
190
−20
−20
−30
−30
1950
1950 2000 2050 2100 2150
Frequency (MHz)
(a)
2000 2050 2100
Frequency (MHz)
2150
(b)
5
4
N.F. (dB)
S 21 (dB)
30
20
10
3
2
1
0
0
1950
2000 2050 2100
Frequency (MHz)
1950
2150
Power consumption (W)
IIP 3 (dB)
30
20
10
0.8
0.6
0.4
0.2
0.1
1950
2000 2050 2100
Frequency (MHz)
(e)
2150
(d)
(c)
0
2000 2050 2100
Frequency (MHz)
2150
1
2
3
Test Volltage (V)
4
(f)
Figure 12.
Measurement performance of improved Sample B
according to different applied voltage. (a) S11 , (b) S22 , (c) S21 ,
(d) noise figure, (e) IIP3, (f) power consumption.
Progress In Electromagnetics Research C, Vol. 22, 2011
(a)
191
(b)
Figure 13. Analysis of LNA (at 4 kV HBM ESD impulse). (a) Before
improved circuit, (b) after improved internal circuit.
Table 5. ESD test result of the improved circuit.
Test Result
Test Voltage
ESD
Test
Voltage
1 kV
2 kV
3 kV
4 kV
A typle
B typle
Number of Number of Number Number of
Sample
Failure
of Sample
Failure
5
0
5
0
5
0
5
0
5
0
5
0
5
0
5
0
directly connected path inside a product, blocking and diverting will
be the answer. Blocking increases impedance of ESD-induced noise
current path, and Diverting changes current’s path so that noise
current would not pass internal circuit. Considering the size, blocking
method shall be more appropriate for a product like LNA.
In this paper, we constructed a circuit with ideal protective diode,
selected from series of tests. By inserting diode to the input, the input
circuit’s impedance has increased, and such increase could block the
sudden flow of excess current induced by static. Figure 10 illustrates
LNA and matching circuit in LNA with protective diode.
The identical ESD test was carried out, following IEC61000-4-2
standard, after the circuit improvement installed. As a result, both
samples A and B showed resistance to at least 4 kV or more. Table 5
shows ESD test result of the improved circuit, and Figures 11 and 12
show each sample’s new measurement property for different applied
voltages. Figure 13 shows analysis of LNA’s improved internal circuit,
observed by ESEM.
6. CONCLUSION
This paper researches on the static effect of communication low-noise
amplifier that utilizes GaAs wafer. It describes the effect of static
on low powered circuit, such as GaAs LNA, and widely used static
192
Kim, Hwang, and Choi
test standard and method, which it implemented to carry out an
experiment on GaAs LNA’s resistance to statics.
ESD test was done on two communication GaAs LNA IC models,
and the result showed that they both had weak resistance to ESD: less
than 2 kV for sample A and less than 3 kV for sample B.
We simulated electro static discharge of GaAs LNA, using
commercial software in order to confirm the amount of static energy
flown into LNA’s internal circuit. Analysis on failure of samples
revealed LNA module’s input/output port had disconnection or
burning problem.
As a countermeasure, appropriate protective diode was installed
on GaAs LNA’s internal circuit, and the improved circuit was proved
to be resistant to more than 4 kV.
REFERENCES
1. Bock, K., “ESD issues in compound semiconductor high-frequency
devices and circuits,” Electrostatic Discharge Symposium, 1–12,
Sep. 23–25, 1997.
2. Howes, M. J. and D. V. Morgan, Reliability and Degradation,
Wiley, New York, 1981.
3. Vinson, J. E. and J. J. Liou, “Electrostatic discharge in
semiconductor devices,” Proc. of the IEEE, Vol. 86, 399–420, 1998.
4. Green, T., “A review of EOS/ESD field failure in military
equipment,” Proc. 10th EOS/ESD Symposium, 1998.
5. Wunsch, D. C. and R. R. Bell, “Determination of threshold
failure levels of semiconductor diodes and transistors due to pulse
voltages,” Nuclear and Plasma Sciences Society, Vol. 15, 244,
Nov. 1968.
6. Schreier, L. A., “Electrostatic damage susceptibility of semiconductor devices,” Reliability Physics Symposium, 151, Apr. 1978.
7. Lacrampe, N., F. Caignet, M. Bafleur, N. Nolhier, and
N. Mauran, “Characterization and modeling methodology for
IC’s ESD susceptibility at system level using VF-TLP tester,”
29th Electrical Overstress/Electrostatic Discharge Symposium, 1,
Sep. 16–21, 2007.
8. Amerasekera, A., W. Abeelen, L. Roozendaal, M. Hannemann,
and P. Schofield, “ESD failure modes: characteristics mechanisms,
and process influences,” IEEE Transactions on Electron Devices,
Vol. 39, 430–436, Feb. 1992.
9. Polgreen, T. L. and A. Chatterjee, “Improving the ESD failure
threshold of silicided n-MOS output transistors by ensuring
Progress In Electromagnetics Research C, Vol. 22, 2011
10.
11.
12.
13.
14.
15.
16.
193
uniform current flow,” IEEE Transactions on Electron Devices,
Vol. 39, 379, Feb. 1992.
Soohoo, K. M. and C.-Y. Wu, “On component level ESD testing,”
Industry Applications Society Annual Meeting IEEE Conference,
872, Oct. 7–12, 1990.
MaCleod, L. M. and K. G. Balmain, “Compact traveling-wave
physical simulator for human ESD,” IEEE Transactions on
Electromagnetic Compatibility, Vol. 39, 89, May 1997.
Zaridze, R., D. Karkashadze, R. G. Djobava, D. Pommerenke, and
M. Aidam, “Numerical calculation and measurement of transient
fields from electrostatic discharges,” IEEE Transactions on
Components, Packaging, and Manufacturing Technology, Vol. 19,
178, Jul. 1996.
Jobava, R., D. Pommerenke, D. Karkashadze, P. Shubitidze,
R. Zaridze, S. Frei, and M. Aidam, “Computer simulation of ESD
from voluminous objects compared to transient fields of humans,”
IEEE Transactions on Electromagnetic Compatibility, Vol. 42, 54,
Feb. 2000.
Giannetti, R. and B. Tellini, “Equivalent network modeling to
simulate electrical discharges,” IEEE Transactions on Magnetics,
Vol. 36, 971, Jul. 2000.
Caniggia, S., “Circuit and numerical modeling of electrostatic
discharge generators,” IEEE Trans. on Industry Applications,
Vol. 42, 1350–1357, 2006.
Smedt, R. D., K. Vervoort, E. V. Oerle, and D. V. Troyen, “Effect
of a board on the shielding effectiveness of a box,” EMC Europe
International Symposium on EMC, 413–418, 2002.